Proceedings Third UJNR Workshop on Soil-Structure Interaction, March 29-30, 2004, Menlo Park, California, USA.

Field Method for Estimating Soil Parameters

for Nonlinear Dynamic Analysis of Single
Piles
A. Anandarajah1, J. Zhang2 and C. Ealy3
Use of in situ soil properties increases the reliability and accuracy of numerical
predictions. The problem of interest here is the nonlinear dynamic behavior of
pile foundations. It is shown in this paper that soil parameters needed for
simplified dynamic analysis of a single pile may be back-calculated from the
dynamic response of the pile measured in the field. A pile was excited by
applying a large horizontal dynamic force at the pile-head level, and the
response measured. In this paper, two different (simplified) methods of
modeling the dynamic response of the pile are considered. One of the methods
is based on the Winkler foundation approach, with the spring constant
characterized by the so-called nonlinear p-y springs. The second method is
based on the equivalent linear finite element approach, with the nonlinearity of
shear modulus and damping accounted for by employing the so-called
degradation relationships. In the latter, the effect of interface nonlinearity is
also considered. Starting with best estimates of soil parameters, the
experimental data on the response of pile is used to fine-tune the values of the
parameters, and thereby, to estimate parameters that are representative of in
situ soil conditions. The soil parameters calibrated by the method can be
applied to earthquake problems when the pore pressure build-up due to freefield response is not very high.
KEYWORDS piles, dynamics, soil, soil-pile system, soil-structure interaction, finite
element method, Winkler method, nonlinear analysis.
INTRODUCTION
The behavior of a deep foundation depends on a set of complex factors such as the
nonlinear constitutive behavior of soils including the effect of pore water pressure, soilpile-superstructure interaction including slip and separation at the pile-soil interface,
characteristics of the loading, superstructure compliance, etc. When the amplitude of
loading is large, most of these factors control the behavior. Accuracy and reliability of the
predicted behavior depend not only on the analysis method employed, but also on the
accuracy with which the model parameters (or soil properties) are determined. In cases
1

Professor,
Graduate student,
Department of Civil Engineering, Johns Hopkins University, Baltimore, MD, 21218, Rajah@jhu.edu
3
Geotechnical Engineer, Federal Highway Administration
2

where good quality undisturbed samples can be obtained, most of the required properties
can be determined by testing them in a laboratory. Due to unavoidable sample
disturbance during sampling, transportation, and preparation during testing, in situ
methods are preferred over laboratory methods. However, heterogeneity of natural soil
deposits and approximations implied in analysis methods cannot be compensated by any
of the above measures. This is, in fact, the reason that the static capacity of piles are in
most cases determined directly from pile load tests performed in the in situ soil. A similar
approach is highly desirable in the design of piles subjected to dynamics loads. It is
shown that parameters for simplified analysis methods such as the Winkler foundation
method and the equivalent linear finite element method may be determined from
measured response of the pile subjected to a large-amplitude dynamic load applied at the
level of the pile head. Experiments were conducted in a uniform granular soil deposit,
filled in a test pit available at the test site, Turner-Fairbank Highway Research center
(TFHRC), Federal Highway Administration, McLean, Virginia.

Fig. 1. A Photograph of the Field Test Setup

EXPERIMENTS
The tests were conducted in a 20 feet deep (6.1m) pit with a plan area of
18 18 (5.5m5.5m). The pit was filled with a uniform sand in loose to medium dense
state about an year prior to the time the tests were conducted (for a different purpose).
During this time period, the sand was subjected to rain several times. At the time the tests
were conducted, the water table was below the level of the pile tip. The sand within the
depth of the pile was damp due to capillary action.
A 4-in (0.1m) diameter, 12.3-feet (3.75m) long, pipe pile was driven into the soil to a
depth of 9.2 feet (2.8m), with an overhang of 3.1 feet (0.95m). A weight of 122 lbs (0.54
kN) was attached to the pile at the pile head. The loading was to be applied by the
Statnamic device (Middendorp, et al., 1992). The Statnamic device produces a singlepulse, impact loading. In order to extract more cycles of vibrations from a single-pulse

impact loading, a spring-mass oscillator was attached to the pile head, and the Statnamic
load was applied in the horizontal direction to the pile at the pile-head level through this
spring-mass oscillator. The test setup is shown in Fig. 1. As a shot is fired from the
Statnamic device, the projectile latches onto the spring (which is attached to the pile
head), and oscillates along with the pile head. Three tests were conducted, each time with

a spring of different spring constant.

Fig. 2. Comparison Normalized Ground-Level Displacement Versus Time Histories of
the Pile From the Three Tests.
The load experienced by the pile head was measured directly by a load cell attached
between the pile head and the excitation setup. The horizontal displacement response of
the pile was measured using 2 LVDTs (Linear Variable Displacement Transformers), one
attached at the pile-head level and the other at the ground level. In addition, the
horizontal pile-head acceleration was measured with the aid of an accelerometer. Fig. 2
presents a comparison of the ground-level horizontal displacement-time histories of the
pile from the three tests. The peak displacements from the 3 tests were 0.8in, 1.5in, and
0.9in (2.0cm, 3.8cm, and 2.3cm) respectively. Except for the differences in the
amplitudes, the frequency responses are almost the same. Only the results of the first test
(Test 1, which has a peak amplitude of 0.8in) are used in the subsequent discussions and
analyses. The load-time history is presented below along with numerical results.
ANALYSES BY THE FINITE ELEMENT METHOD
Details of the Method
The analyses presented in this paper are performed using HOPDYNE (Anandarajah,
1990), which is a finite element computer program with capabilities to model soil (linear
and nonlinear), foundations, superstructure (linear) and soil-structure problems. The
primary features of HOPDYNE that are used in the present study are

8-noded solid isoparametric element to model the soil

2-noded beam bending element to model the pile
4-noded three-dimensional slip elements to model the soil-pile interface
Equivalent-linear approach to account for the nonlinearity of the soil

Details of these features can be found in published literature, and will not be repeated
here.
The soil properties needed for equivalent linear finite element analysis are (for each soil
type): Gmax , min , G versus eff relation and versus eff relation. The pile is
characterized in terms of I xx , I yy , J zz , A p , E and , where I xx , I yy and J zz are the
second moment of inertias about x , y and z axes respectively (with the pile axis
taken along the z axis), A p is the cross sectional area of the pile, E is the Youngs
modulus and is the Poissons ratio. In addition, mass densities of the soil and pile are
also required. While the properties of the pile are known and fixed, the properties of the
soil are to be back calculated. To initiate the iterative process, starting estimates are
required. The empirical equation suggested by Seed and Idriss (1970) is used to obtain a
starting estimate for G max :
G
= K 1 / 2
(1)
max

where K is a density-dependent constant, and m is the mean normal pressure. As the

value of the coefficient of earth pressure at rest is unknown, it is assumed to be 1.0. With
this assumption, m becomes equal to the vertical effective stress. For a loose sand, with
m expressed in lb/ft2 (or psf), K = 40,000 . The density of the sand was about 110 lb/ft3
(17.3 kN/m3). As the soil is homogeneous, and the water table is not within the depth of
the pile, Eq. 1 becomes:
G max = 40,000 (110h )1 / 2 = 0.42 10 6 h1 / 2 psf

(2)

where h is the soil depth. The optimal value of Gmax is then taken as
G = F G

(3)
where F1 is a multiplier, to be established iteratively by matching the experimental
response to the theoretical response.
max

max

The value of min was found to have a negligible influence on the overall response. On
this basis, a value of 0.005% is assumed in the analyses reported here.
The G versus eff relation and versus eff relation are function of the soil types. As
the soil type is known, the empirical relations proposed for this soil type (sand) by Seed
and Idriss (1970) are used, and assumed to be fixed. Thus, the only parameter that is
sought by the back-calculation process is F1 .

Fig. 3. Deformation of a Portion of the Domain Analyzed

Analysis and Results

The problem of interest is complex and three-dimensional. Three-dimensional finite
element analyses can be computationally intensive, even in the case of a total stress based
equivalent linear analysis such as the one of interest here. Equivalent linear analyses
amount to a few (3 to 10) linear analyses. When pore water pressure effects are to be
considered (which is the natural extension of total stress analyses), the computational
efforts can be overwhelming to the point where the analyses can no longer be performed
within a few minutes on a PC. Thus, from the point of view of using the analyses for
practical design purposes, it is of interest to explore approximations that may yield results
with acceptable accuracy.

Fig. 4. Plain-Strain Approximation of Gazetas and Dobry (1984)

First, we consider a regular mesh with no approximations. As the problem is symmetric

about the vertical plane that contains the pile and the direction of applied loading, only
half the domain needs to be discretized. This cylindrical mesh, which contains 2824
nodes, 2240 soil elements, and 18 bending elements, provides accurate results for the
problem. The outer boundary of the cylindrical domain is placed at a radius of 30 feet
(9.15m). There are 14 layers of elements in the vertical direction. The deformation is
pretty much confined to the region near the pile. To show the details a little better,
deformation of the domain (at a given time) within a window around the pile is shown in
Fig. 3.
Then two specific approximations are considered. In the past, several approximations
have been developed for obtaining analytical solutions in the frequency domain (e.g.,
Gazetas and Dobry, 1984) and for approximate finite element solutions (e.g., Wu and
Finn, 1997). We will use some of these as guidelines for our finite element analyses. In
particular, Gazetas and Dobry (1984) developed models based on plain-strain
approximations. In this, a slab of soil, extending to infinity in the radial direction, is
assumed to have no displacements in the direction normal to the plane of the slab. At the
center of the slab is a square-shaped pile segment undergoing a horizontal dynamic
motion. The slab is divided into four quarters. As shown in Fig. 4, the energy that radiates
away from the pile into the soil is assumed to take place in two distinct ways: (1) in the
form of a compressional wave through two of the quarters, and (2) in the form of a shear
wave through the other two quarters. It was shown that the radiation damping calculated
from this approximate plane-strain model closely matched that radiation damping
calculated from the plane strain solution of Novak, et al. (1978). The 14 plain-strain
slabs, each with 4 quarters, are attached to the pile. In the radial direction, the slabs are
fixed at 30 feet (0.915m) from the center. The slabs are not bonded in the vertical
direction; i.e., each slab can undergo horizontal motions independently of each other.

Fig. 5. Deformed Configuration Using Approximate, Coupled Mesh

Disp.-Pile at G-Level (in)

The effect of bonding the plain-strain slabs in the vertical direction is examined using the
mesh shown in Fig. 5, where, owing to symmetry, only half the domain is discretized.
Primary difference between the coupled and uncoupled meshes are that the horizontal
slabs are not bonded (i.e., uncoupled) to each other in the uncoupled mesh whereas they
are in the coupled mesh. To further cut down the number of degrees of freedom, the
vertical degree of freedom was suppressed in all of the analyses presented; its effect is
examined next.
uz=0, XO=30'
uz=0, XO=15'
Non-zero uz, XO=15'

0.4
0.2
0
-0.2
0

0.5

t (sec)

Fig. 6. Effect of Radial Domain Size and Vertical Degree of freedom

Fig. 6 presents a comparison of ground-level displacement-time histories obtained with
the full mesh under the following conditions: (1) X 0 = 30 and u z = 0 , (2) X 0 = 15 and
u z = 0 and (3) X 0 = 15 and u z 0 , where X 0 is the radial distance at which a fixed
vertical outer boundary is placed, and u z s are the vertical degrees of freedom. It is seen
that the differences in the results are very slight, indicating that the vertical response of
the soil and the pile is negligible in the present case, and the radial distance of 30 feet is
far enough to place a fixed vertical boundary.
Full mesh
Plane strain mesh
560 elem.
coupled mesh

Disp.-Pile at G-Level (in)

0.8
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8

0.5

t (sec)

Fig. 7. Comparison of Results From Full, Uncoupled Plain-Strain and Coupled Meshes
Fig. 7 presents a comparison of ground-level displacement-time histories obtained with
the three different meshes described earlier: (1) the full mesh, (2) the plain-strain mesh,
and (3) the coupled mesh. While the results are not equal to each other, the approximate
models appear to give acceptable results for practical use.

500

P(t) (lbs)

0
-500
-1000

Measured

-1500
-2000
0

0.5

t (sec)

Measured
a(t) (g)

10
5
0
-5
0

t (sec)

Calculated
FE (560 elem) with 0.5Gmax

10
a(t) (g)

0.5

5
0
-5
0

0.5

Measured
Calculated
FE (540 elem) with 0.5Gmax

0.75
Disp.-Pile at G-Level (in)

t (sec)

0.5
0.25
0
-0.25
-0.5
0

0.5

t (t)

Fig. 8. Comparison of Calculated (Using Full Mesh and 0.15Gmax ) and Measured Results

It should be noted that these are not general results, and the outcome might differ from
problem to problem, depending on the frequency of loading, natural frequencies of the
system, etc. However, the results do indicate that for a given problem, it is worth
exploring these approximations so that subsequent analyses (e.g., parametric study or a
8

more systematic probabilistic study, where the analyses need to be repeated several
times) may be performed using one of these approximate models.
After a few trial runs, a reasonably good match is found for F1 = 0.15 . A comparison
between the finite element results (using the full mesh) with F1 = 0.15 (Eq. 3) and the
experimental results are shown in Fig. 8. The first plot in Fig. 8 presents the measured
pile-head load versus time history, which is used as input to the finite element analysis.
The 2nd and 3rd plots present the measured and calculated pile-head acceleration time
histories respectively. The 4th plot presents a comparison between the measured and
calculated ground-level displacement time histories. All of the above quantities are
horizontal components.
In view of the fact that the soil is actually an elasto-plastic material, whereas it is
represented by a form of a nonlinear viscoelastic model in the finite element analysis, the
quantitative comparison shown in Fig. 8 is considered to be reasonably good. In both
cases, the response dies out in about 3 cycles. The rate of decay of the displacement
amplitude is predicted reasonably well.
560 elem. coupled mesh
560 elem. coupled mesh
with slip elements

Disp.-Pile at G-Level (in)

0.8
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8

0.5

t (sec)

Fig. 9. Comparison of Calculated Responses with and without Slip Elements and
0.15Gmax
560 elem. coupled mesh
560 elem. coupled mesh
with slip elements

Disp.-Pile at G-Level (in)

0.8
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8

0.5

t (sec)

Fig. 10. Comparison of Calculated Responses with and without Slip Elements and
0.20Gmax
In the analyses described so far, the soil has been bonded to the pile, preventing any
gapping or slipping to take place at the soil-pile interface. Using the 560-element coupled
mesh (Fig. 5), an analysis is conducted with slip elements placed between the pile

elements and the soil elements. The calculated ground-level displacement time histories
with and without the slip elements are shown in Fig. 9. It is noted that the use of slip
elements renders the response softer, yielding larger pile displacements. These results
were obtained with F1 = 0.15 . Then F1 is varied until a good match is obtained. The
numerical results with F1 = 0.20 and with slip elements are compared in Fig. 10 with
numerical results using F1 = 0.15 and without slip elements. The results from the both
analyses are virtually identical, indicating that the soil stiffness doesnt need to be
reduced as much when slip elements are employed, since the use of slip elements makes
the system softer by allowing slip and separation at the pile-soil interface. It should be
noted that the consequence of allowing slip and separation may be more dramatic in other
problems, and thus having the capability to use slip elements is desirable.

WINKER FOUNDATION METHOD

Details of the Method

Fig. 11. Winkler Foundation Approach: Springs and Dashpots to Represent the Effect of
Soil
In the Winkler foundation approach (Fig. 11), the primary member analyzed is the pile.
The influence of the surrounding soil on the pile is introduced through a series of
nonlinear springs (in the case of static problems). The most widely used spring forcedisplacement relationships are the so-called p-y curves of Matlock (1970) for clays and
Reese, et al. (1974) for sands. In extending this static method to problems involving
dynamic loads such as that from earthquakes, ship collisions, etc., methods are needed for
accounting for damping both material and radiation damping. Several researchers have
worked on this problem (Kagawa, 1980; Berger, et al., 1977; Wang, et al., 1998;
Boulanger at al., 1997; Loch et al., 1998; Badoni and Makris, 1996; Abghari and Chai,
1995, Nogami , et. al., 1992, Gazetas and Dobry, 1984, Sen et al., 1985; Trochanis, et al.,
1991). These studies indicated that there are some major difficulties to be resolved.

10

Firstly, the most appropriate spring/dashpot model to use is not clear. For instance,
Wang, et al. (1998), after comparing results with series (spring and dashpot in series) and
parallel (spring and dashpot in parallel) models, determined that the parallel model leads
to a very stiff system, with the damping force over-dominating the system response. In
either case, the material damping needs to be considered as well. The appropriate model
to use is thus yet to be identified.
Secondly, the issue concerning a suitable value to use for the coefficient of damping has
not been resolved. For instance, if one uses Bergers model (1977) to represent the
radiation damping, where it is assumed that radiation of energy away from the pile takes
place in the form of p- and s-waves through a volume of soil of constant cross section
(like a one-dimensional rod), the damping coefficient becomes frequency independent,
and is given by

C = C p + C s = (V p + Vs ) d

(4)

where V p is the p -wave velocity and Vs is the shear (s-) wave velocity, is the mass
density and d is the diameter of the pile. Here C is the coefficient of damping per unit
length of the pile. Wang et al. (1998) found that the value of C calculated using the above
equation was too large. They arbitrarily assumed C = 2Vs d . There is no consensus
among researchers as to the most suitable model and the most appropriate equation for
computing C. Bergers model is approximate. The manner in which the energy radiates
away from the pile is complex, and the p- and s-wave portions cannot be easily separated
out as is done in Bergers model. The cross section of the portion of the soil that carries
the radiation energy is not constant. But analyses with non-uniform cross sections lead to
frequency-dependent damping parameters (e.g., Gazetas and Dobry, 1984), making it
difficult to apply them in a time-domain analysis. A frequency has to be arbitrarily
selected for computing a value for C to use in a time-domain analysis such as that
involved in the beam of nonlinear Winkler foundation method (BNWF). When the pile is
shaken with a large amplitude loading the problem of interest here - the material
damping is more important than the radiation damping (Brown and ONeill, 2001), and
there is no rational method of calculating a value for the damping coefficient.
There are yet other factors that cannot be properly accounted for at the present time. For
instance, the softening that takes place at the soil/pile interface due to slipping and
gapping is difficult to model. Also, the pore water pressure build up during a cyclic
loading such as the earthquake loading cannot be accurately modeled. All of the above
difficulties associated with the use of nonlinear Winkler foundation methods point to the
need for a site-specific, field calibration of the method.
In the specific Winkler foundation model used here, the soil is replaced by a series of
elements involving a spring and a dashpot in parallel (i.e., without a second series
dashpot shown in Fig. 11). The coefficient of damping is calculated according to Eq. 4,
but with a modifier Fc , as follows:
C = Fc (C p + C s ) = Fc (V p + Vs ) d
(5)

11

Estimated values of V p and Vs are 738 ft/s (225 m/s) and 492 ft/s (150 m/s), and the
mass density is 3.42 lb-sec2/ft4 (1.76 kN-sec2/m4 or 1.76 g/cm3).
The spring constant is represented by the p-y relation suggested by Reese, et al. (1974)
for sands. Parameters of the p-y relation depend on whether the loading is static or cyclic;
here the cyclic parameters are used. The entire curve is a function of ( , , d , k1 ) , where
is the friction angle of the soil, is the effective density of the soil, d is the diameter
of the pile, and k1 is the slope of the initial straight line. The value recommended for k1
for loose to medium dense sand is 60 lb/in3 (16225 kN/m3). Except for d , all other
parameters can assume different values than our initial estimates. Let us, therefore,
introduce multipliers with each parameter as:

= F *

(6a)

= F *

(6b)

k1 = Fk k1*

(6c)

The estimated value of the friction angle for this soil is 350, and the density is 110 lb/ft3
(17.3 kN/m3).

Analysis and Results

HOPDYNE (1990) has the capability to consider nonlinear, discrete springs and dampers.
The beam bending elements available in HOPDYNE is used to model the pile.
The soil deposit is divided into 14 layers, and a parallel spring/dashpot elements is
attached to the pile in the middle of each of these 14 layers. After several trials with
different values for Fc , F , F and Fk , it is found that most matching results are obtained
with Fc = 1, F = 0.5, F = 1 and Fk = 1 . In other words, changing only the value of not
only is adequate, but gives the most optimal results. It may, however, be necessary to
change some of the other parameters for best results in other problems.
The comparison between numerical and experimental results is presented in Fig. 12,
where the plots at the top and middle are the measured and calculated pile-head
horizontal acceleration-time histories, and the plot at the bottom is a comparison between
numerical and experimental ground-level horizontal displacement-time histories. It is
seen that the comparison is as good as the one with the finite element results (Fig. 8),
indicating that both the equivalent linear finite element method and the beam of nonlinear
Winkler foundation method are equally capable of representing the dynamic response of
the single pile under study here.

12

Acce.-Pile Top (g)

Measured

10
5
0
-5

Acce.-Pile Top (g)

0.5 t (sec)

Caculated by BNWF

10
5
0
-5
0

0.5

t (sec)

Disp.-Pile G-Level (in)

Measured
Calculated by BNWF
0.6
0.4
0.2
0

-0.2
-0.4
-0.6
0

0.5

t (sec)

Fig. 12. Comparison Between Calculated (by BNWF Method with F = 0.5 ) and
Measured Results
CONCLUSIONS
A series of large amplitude dynamic tests was conducted on a single pile driven into a
homogeneous sandy deposit. The pile was subjected to a horizontal impact load with the
aid of a Statnamic device. The pile underwent a cyclic motion involving about 3 cycles.
The pile-soil system was modeled using two different numerical methods: (1) Equivalent
linear finite element method, and (2) beam of nonlinear Winkler foundation method with
nonlinear p-y curves to represent the stiffness and Bergers model to represent the
damping. The objective was to back-calculate from the experimental results site-specific
soil properties. The study indicates that both numerical methods are equally capable of
representing the nonlinear dynamic response of the pile-soil system, and that the relevant
soil parameters for these methods may indeed be back-calculated from the experimental
data. While the beam of nonlinear Winkler foundation is simpler and computationally

13

more efficient than the finite element method, the latter has the advantage of having the
capacity to consider the interface slip and separation, and the effect of pore pressure (not
considered in the present study).
ACKNOWLEDGEMENT
The research was supported by the U.S. National Science Foundation under grant No.
CMS0084899 and by the Federal Highway Administration. The cognizant program
director of NSF is Dr. Richard Fragaszy. These supports are acknowledged.
REFERENCES
Abghari, A. and Chai, J. (1995). ``Modeling of soil-pile superstructure interaction in the
design of bridge foundations,'' Geotechnical Specialty Pub. No. 51, Performance of Deep
Foundations Under Seismic Loading, ASCE, John Turner (ed.), 45-59.
Anandarajah, A. (1990). HOPDYNE: A finite element computer program for the analysis
of static, dynamic and earthquake soil and soil-structure systems. Civil Engineering
Report, Johns Hopkins University.
Berger, E., Mahin, S.A. and Pyke, R. (1977). Simplified Method for Evaluating SoilPile Structure Interaction Effects. Proc. 9th. Offshore Technology Conference, OTC
Paper No. 2954, Huston, Texas, pp. 589-598.
Boulanger, R.W., Wilson, D.W., Kutter, B.L., and Abghari, A. (1997). ``Soil-Pile
Superstructure Interaction in Liquefiable Sand.'' Transportation Research Record 1569.
Brown, D. A., and ONeill, M.W. (2001). Static and Dynamic Response of Pile
Groups. NCHRP Study (report not available yet).
Gazetas, G. and Dobry, R. (1984). Simple Radiation Damping Models for Piles and
Footings. J. Engrg. Mech., ASCE, 110(6), pp. 937-956.
Hilber, H.M., Hughes, T.J.R. and Taylor, R.L. (1977). Improved numerical dissipation
for time integration algorithms in structural mechanics. Int. J. Earthquake Eng. And
Structural Dynamics. 5:283-292.
Idriss,, I.M., Lysmer, J., Hwang, R. and Seed, H.B. (1973): Quad4: A computer program
for evaluating the seismic response of soil structures by variable damping finite element
procedure, University of California, Berkeley Report No. EERC 73-16.
Kagawa, T. (1980).``Soil-Pile Structure Interaction of Offshore Structures During an
Earthquake.'' Proc. 12th. Annual Offshore Technology Conference, Huston, Texas, OTC
3820.

14

Matlock, H. (1970). Correlations for Design of Laterally-Loaded Piles in Soft Clay.

OTC Paper No. 1204, Huston, Texas.
Matlock, K., Foo, S.H. and Bryant, L.L. (1978). Simulation of Lateral Pile Behavior.
Proc. Earthquake. Engrg. Soil Dyn. ASCE, July, pp. 600-619.
Middendorp, P., Bermingham, P. and Kuiper, B. (1992). Statnamic Load Testing of
Foundation Piles. Proc. 4th. Intl. Conf. Application of Stree-Wave Theory to Piles, The
Hague, The Netherlands, pp. 581-588.
Novak, M., Nogami, T. and Aboul-Ella, F. (1978). Dynamic soil reactions to pile
vibrations. J. Eng. Mechanics Division, ASCE, 104(EM4):1024-1041.
Nogami, T., Otani, J., Konagai, K. and Chen, H-L. (1992). ``Nonlinear Soil-Pile
Interaction Model for Dynamic Lateral Motion.'' J. Geotech. Eng., ASCE, 118(1):89106.
Reese, L.C., Cox, W.R., and Kooper, F.D. (1974). ``Analysis of Laterally-Loaded Piles in
Sands.'' Proc. 6th. Annual Offshore Technology Conference, Huston, Texas, OTC 2080.
Seed, H.B. and Idriss, I.M. (1970). Soil Moduli and damping factors for dynamic
response analyses. University of California, Berkeley Report No. EERC 73-16.
Sen, R., Davis, T.G. and Banerjee, P.K. (1985). Dynamic Analysis of Piles and Pile
Groups Embedded in Homogeneous Soils. Intl. J. Earthquake Engrg. Soil Dyn., 13:5365.
Trochanis, A.M., Bielak, J. and Christiano, P. (1991). Simplified Model for Analysis of
One or Two Piles. J. Geotech. Engrg., ASCE, 117(3):448-466.
Wang, S., Kutter, B.L., Chacko, M.J., Wilson, D.W. and Boulanger, R.W. (1998).
``Nonlinear Seismic Soil-Pile Structure Interaction.'' Earthquake Spectra, 14(2):377-396.
Wu, G. and Finn, W.D.L. (1997). Dynamic nonlinear analysis of pile foundations using
finite element method in the time domain. Canadian Geotechnical J. 34:44-42.
Zienkiewicz, O.C. and Taylor, R.L. (1989). The finite element method. Vol I, 4th Edition,
MacGraw-Hill.

15